Twin mapping method integrating power mode and deep learning model

Abstract

The embodiment of the present invention discloses a twin mapping method that integrates a dynamic mode and a deep learning model, including: collecting environmental factor data of a target area in real time according to a twin mapping scenario to be simulated; using a physical dynamic mode to numerically simulate the twin mapping scenario of the target area, and using the environmental factor data to correct the numerical simulation results to obtain spatial distribution data of the environmental factor; using the spatial distribution data of the environmental factor as a sample to train a lightweight deep learning model, the trained model uses the historical spatial distribution data of the environmental factor as input and the future spatial distribution data of the environmental factor as output; inputting the environmental factor data collected in real time into the trained model, dynamically predicting the future environmental factor data of the target area in real time and displaying it visually. This embodiment provides twin mapping and visualization of the entire life cycle of the geodynamic process.

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of digital twins, and in particular to a twin mapping method that integrates a power mode and a deep learning model. Background Art

[0002] Geographic Information System (GIS) is the digitization of physical geographic systems. It maps physical geographic systems into computers in the form of points, lines, surfaces, and volumes, making them information geographic systems. With the development of sensor networks, the Internet of Things, Building Information Modeling (BIM), modern communications and other technologies, GIS has developed into digital twins. Digital twins were first used by NASA in the Apollo program in 1969 to build twins of spacecraft. In 2003, the concept of digital twins was formally proposed by Michael Grieves of the University of Michigan. It was initially named the "Information Mirroring Model" and later evolved into the term "digital twin". To date, many different definitions of digital twins have been born, but in essence, they can be understood as "3D clones" or "twins" of physical space, which are twins of static physical space.

[0003] However, the geographical process of physical space changes with time, scene, granularity, etc. For example, the self-purification process of pollutants exchanged between island reef lagoons and offshore waters will change with time, ocean currents, monsoons, etc. The self-purification process of water quality in lagoons has the characteristics of dynamic physical processes. How to twin map and visualize the dynamic geographical process of physical space is an urgent problem to be solved. Summary of the invention

[0004] An embodiment of the present invention provides a twin mapping method that integrates dynamic patterns and deep learning models, which can quickly calculate spatial distribution field data at different times in the future based on real-time monitoring data, and realize twin mapping and visualization of the entire life cycle of geographic dynamic processes.

[0005] In a first aspect, an embodiment of the present invention provides a twin mapping method integrating a power mode and a deep learning model, including:

[0006] According to the twin mapping scenario to be simulated, the environmental factor data of the target area is collected in real time;

[0007] Using the physical dynamic model, numerically simulate the twin mapping scene of the target area, and use the environmental factor data to correct the numerical simulation results to obtain the spatial distribution data of the environmental factor;

[0008] The spatial distribution data of the environmental factors are used as samples to train a lightweight deep learning model, wherein the trained model uses the historical spatial distribution data of the environmental factors as input and uses the future spatial distribution data of the environmental factors as output;

[0009] The environmental factor data collected in real time is input into the trained model, and the future environmental factor data of the target area is predicted in real time and displayed visually.

[0010] In a second aspect, an embodiment of the present invention provides an electronic device, the electronic device comprising:

[0011] one or more processors;

[0012] a memory for storing one or more programs,

[0013] When the one or more programs are executed by the one or more processors, the one or more processors implement the twin mapping method of the fusion power mode and the deep learning model described in any embodiment.

[0014] In a third aspect, an embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, which, when executed by a processor, implements the twin mapping method of the fusion power mode and the deep learning model described in any embodiment.

[0015] The embodiment of the present invention provides a twin mapping method that integrates physical dynamic mode and deep learning model. It uses sensor real-time monitoring technology, combines traditional physical dynamic mode and lightweight deep learning model, and can quickly calculate the spatial distribution field data at different times in the future according to real-time monitoring data, so as to realize the twin mapping and visualization of the whole life cycle of geodynamic process. The present invention overcomes the shortcomings of traditional high-precision hydrodynamic mode, such as large amount of calculation, slow calculation speed, and dependence on large supercomputers for computing power. It can quickly calculate the spatial distribution of pollutant concentration field and realize the rapid twin mapping and visualization of geographic process. However, the sensor monitoring data can only represent the "instantaneous" state of water environment elements at a certain monitoring point, and it is difficult to reflect the large-scale spatial distribution changes of geodynamic process. The present invention can quickly calculate the spatial distribution field data of island, reef and lagoon water environment at different times in the future according to real-time monitoring data and using lightweight deep learning model. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 It is a flow chart of a twin mapping method integrating a power mode and a deep learning model provided by an embodiment of the present invention;

[0018] Figure 2 It is an architecture diagram of a ConvLSTM model provided by an embodiment of the present invention;

[0019] Figure 3 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.

[0021] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0022] In the description of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0023] Figure 1 This is a flow chart of a twin mapping method that integrates a power mode and a deep learning model provided by an embodiment of the present invention. The method is applicable to the case of simulating a twin mapping scenario with a dynamic physical process and is executed by an electronic device. Figure 1 As shown, the method specifically includes:

[0024] S110. Collect environmental factor data of the target area in real time according to the geodynamic process of the twin mapping scene.

[0025] The twin mapping scene here refers to the geodynamic scene to be simulated, and different twin mapping scenes require different environmental data. This embodiment first determines the actual environmental data to be collected according to the geodynamic process of the specific scene as the data source of the entire method.

[0026] In a specific implementation, taking the twin mapping scenario of the self-purification of pollutants in island reef lagoons as an example, the corresponding target area is the lagoon, and the geodynamic process in the scenario is the dynamic process of water exchange between the lagoon and the open sea. First, according to the dynamic process of water exchange between the lagoon and the open sea in the self-purification of pollutants in island reef lagoons, relevant sensors are deployed near the sewage outlets of island reef lagoons. Then, the water environment factor data of the lagoon is collected in real time through each sensor, including pollutant concentration (such as nitrogen concentration, phosphorus concentration), seawater temperature, seawater salinity, ocean current field and other water environment element data.

[0027] S120. Use the physical dynamic model to perform numerical simulation on the twin mapping scene of the target area, and use the environmental factor data to correct the numerical simulation results to obtain the spatial distribution data of the environmental factors.

[0028] Still taking the self-purification scenario of island reef lagoon pollutants as an example, this step uses mature physical dynamic models, such as FVCOM (Finite Volume Community Ocean Model, a high-precision hydrodynamic exchange model), to numerically simulate the lagoon water exchange process, and correct its parameters in combination with real-time monitoring data to ensure that the simulation results are accurate and reliable. This process may include the following steps:

[0029] Step 1: Determine the model parameterization scheme based on the opening and closing of the island and reef lagoon, terrain and wind field effects. Optionally, set the large-scale sea area where the island and reef lagoon is located as the initial area to determine the parameterization scheme for the model operation, which includes the simulation time resolution, spatial grid resolution, simulation time range, lagoon bottom type, lagoon mixing rate (mixing rate of temperature and salinity inside the lake), etc.

[0030] Step 2: Based on the historical observation data and field survey data of island and reef lagoons, the physical dynamic model is used to numerically simulate the lagoon water exchange process. Optionally, four types of external forcing conditions, such as wind, open boundary, river, and meteorological conditions, are first used to run the large-scale coarse-resolution FVCOM hydrodynamic model to simulate the large-scale field data of water environmental factors, namely nitrogen concentration, phosphorus concentration, seawater temperature, seawater salinity, current field and other large-scale field data. Then, according to the shoreline, topography and geomorphic data of the target island and reef lagoon, the island and reef lagoon area is refined and nested in the network to construct the FVCOM high-precision hydrodynamic model of the island and reef lagoon area. Thirdly, the output results of the large-scale coarse-resolution FVCOM hydrodynamic model are used as open boundary conditions for nitrogen concentration, phosphorus concentration, seawater temperature, seawater salinity, and ocean current field. Combined with the observation data of temperature, salinity, ocean currents, etc. in the historical period of island and reef lagoons and the survey data of ocean dynamic environment such as wind, waves, and currents, the FVCOM high-precision hydrodynamic model of the island and reef lagoon area is called to simulate and generate water environment data of the lagoon in the historical period.

[0031] Step three, use the environmental factor data collected in real time to correct the parameters of the numerical simulation results to ensure that the simulation results are accurate and reliable, and finally obtain the spatial distribution data of nitrogen concentration, phosphorus concentration, seawater temperature, seawater salinity, and ocean current field at different times in the lagoon. Optionally, the environmental factor data collected in real time can be interpolated and preprocessed to convert it into data in the same format as the numerical simulation results; then use the data assimilation method to merge the converted data with the numerical simulation results to obtain the spatial distribution data of nitrogen concentration, phosphorus concentration, seawater temperature, seawater salinity, and ocean current field at different times in the lagoon, wherein the data assimilation method includes at least one of the variational method, Kalman filtering, or particle filtering, and the specific process of each data assimilation is prior art and will not be repeated here.

[0032] Since the small area of ​​the island reef lagoon is not completely closed, but is affected by the external ocean temperature field, salinity field, tide and other conditions, long-term data observation is required to obtain these conditions. When it comes to a specific island reef, it is usually difficult to obtain the required data. Therefore, this embodiment constructs a larger-scale coarse-resolution dynamic model of the sea area where an island reef is located in the above-mentioned manner. Due to its low resolution requirements and large simulation range, it is relatively easy to find the basic data required for model construction and provide the target island reef lagoon with the required basic boundary condition data. If a designated small area has been observed for a long time and all kinds of data are complete, there is no need to build a large-scale model, and high-precision modeling can be directly performed on the small area. This embodiment does not make specific restrictions.

[0033] S130. Using the spatial distribution data of the environmental factors as samples, a lightweight deep learning model is trained. The trained model uses the historical spatial distribution data of the environmental factors as input and the future spatial distribution data of the environmental factors as output.

[0034] Physical dynamic mode technology has disadvantages such as large amount of calculation, slow operation speed, and heavy reliance on supercomputers. Simply using this technology for twin mapping of geographic processes cannot meet the real-time requirements. In order to make up for this defect, this embodiment introduces a deep learning model with extremely high computational efficiency into the twin mapping of geographic processes, inputs a large amount of time-series geographic process data into the deep learning model for training, obtains the evolution law of geographic processes from the data, and realizes rapid simulation of geographic processes, thereby meeting the needs of real-time mapping of geographic processes. Optionally, a convolution-long short-term memory neural network model (ConvLSTM) can be used as the lightweight deep learning model, and the spatial distribution data of environmental factors at multiple consecutive moments before and after any moment are extracted as samples; the spatial distribution data of environmental factors at multiple consecutive moments before the moment are input into the lightweight deep learning model for training, so that the model output continuously approaches the spatial distribution data of environmental factors at multiple consecutive moments after the moment.

[0035] Still taking the self-purification scenario of island reef lagoon pollutants as an example, first, based on the simulation data of high-precision hydrodynamic exchange patterns obtained in S120 (spatial distribution data of seawater temperature, salinity, current field, nitrogen concentration, phosphorus concentration, etc. at different times in the lagoon), training samples are constructed and divided into training set, validation set and test set in a ratio of 7:2:1. The ConvLSTM model is a supervised learning model, which needs to determine the input features and output labels. The determination method is to select T 1 ,T 2 ,…,T n , that is, the field data at n moments is used as the input feature, and T n+1 ,T n+2 ,…,T n+k , that is, the field data at k moments is used as the output label. This method is applied to the training set, validation set, and test set to divide the features and labels. Then, the ConvLSTM model is trained using the above data set, and the deep learning model parameters are gradually optimized through the gradient descent method to minimize the training set error. Finally, the validation set and test set are used to evaluate the model effect. If the result does not meet the accuracy requirements, the hyperparameters need to be changed and the model needs to be retrained until the model accuracy meets the requirements.

[0036] Furthermore, one can use Figure 2The ConvLSTM model architecture shown in Figure 1. The model is divided into two layers, and deeper layers can better extract the feature information of the data. The model input data X(t, input_channels, height, weight) and output data Y(t, output_channels, height, weight) are both 4-dimensional tensors, where the first parameter t represents the time step, the second parameter represents the number of channels of the input and output grid data, and each channel represents a variable, such as temperature, salinity, etc. The third and fourth parameters represent the height and width of the input grid data.

[0037] S140, inputting the environmental factor data collected in real time into the trained model, dynamically predicting the future environmental factor data of the target area in real time and displaying them visually.

[0038] In the example of the self-purification process of pollutants exchanged between island reef lagoons and offshore water quality, the water environment element data of island reef lagoons monitored in real time can be used as input, and a lightweight deep learning model can be called to simulate and calculate the spatial distribution data of the lagoon water environment at different times in the future, such as the spatial distribution data of the water environment in the next 1 hour, 2 hours, 3 hours, ..., 6 hours (grid field data), and store them in a fixed file directory. When visualizing, extract the environmental factor data in a continuous period from the fixed file directory; for the grid field data in the environmental factor data, use gradient colors to perform hierarchical visualization in chronological order, wherein the grid field data includes nitrogen concentration, phosphorus concentration, seawater temperature and seawater salinity; perform particle flow visualization on the ocean current field data; and superimpose the ocean current field data visualization layer on the grid field data visualization layer to realize the dynamic visualization of the nitrogen concentration, phosphorus concentration, seawater temperature and seawater salinity in the lagoon as the ocean current field changes.

[0039] In one embodiment, the data processing and visualization display process can be completed by the background program and the front-end program of the network service system, respectively. The background program periodically reads the 12-hour water environment spatial distribution field data of the island reef lagoon from a fixed file directory (centered on the data at the current moment, the data of the first 5 moments and the last 6 moments), and sends the water environment spatial distribution field data to the front-end program. After receiving the data, the front-end program uses gradient colors to perform hierarchical visualization of the raster field data such as seawater temperature, salinity, nitrogen concentration, phosphorus concentration, etc. in the lagoon in chronological order; and uses a particle flow visualization method for the vector data of the current field in the lagoon. The particle flow visualization of the current field is superimposed on the dynamic visualization layer of the raster field data to realize the dynamic visualization of the concentration of pollutants such as nitrogen and phosphorus in the lagoon as the current field changes.

[0040] As time goes by, the sensor acquires new real-time monitoring data and continuously repeats the S140 data prediction and dynamic display process, dynamically displaying on the front end the dynamic changes in the concentration of pollutants such as nitrogen and phosphorus in the lagoon as the ocean current field changes.

[0041] The remaining parts not described in detail belong to the common knowledge of those skilled in the art.

[0042] It is worth mentioning that the traditional physical dynamic mode technology mainly simulates and generates field data at different times based on historical observation data and field survey data, but this technology has disadvantages such as large amount of calculation, slow operation speed, and heavy reliance on supercomputers. When this technology is used alone for twin mapping of geographic processes, it cannot meet the real-time requirements. In order to make up for this defect, this embodiment introduces a deep learning model with extremely high computing efficiency into the twin mapping of geographic processes, inputs a large amount of time-series geographic process data into the deep learning model for training, obtains the evolution law of geographic processes from the data, and realizes the rapid simulation of geographic processes, thereby meeting the needs of real-time mapping of geographic processes. However, the deep learning model has high requirements for the quality and amount of training data. Generally, it is difficult to have high-precision observation data of a long time series for a geographic process in a small range of scenes. To this end, this embodiment integrates the physical dynamic mode with the deep learning model, relies on the physical dynamic mode to generate time-series high-precision geographic process simulation data, and provides it as a training data set to the deep learning model, which not only makes up for the data missing problem, but also can meet the needs of real-time mapping of geographic processes. In addition, since both the physical dynamic model and the deep learning model are simulations of geographic processes, in order to prevent error accumulation and ensure that the errors between the simulation results and the real scene data are within the allowable range, this embodiment deploys sensors for the target geographic process to obtain real-time monitoring data, and uses the real-time monitoring data to correct the above-mentioned model parameters to ensure that the simulation results are accurate and reliable, and ultimately meet the real-time and authenticity requirements of the twin mapping of the geographic process.

[0043] In summary, this embodiment provides a twin mapping method that integrates physical dynamic mode and deep learning model. It uses sensor real-time monitoring technology, combines traditional physical dynamic mode and lightweight deep learning model, and can quickly calculate the spatial distribution field data at different times in the future based on real-time monitoring data, so as to realize the twin mapping and visualization of the entire life cycle of the geodynamic process. The present invention overcomes the shortcomings of traditional high-precision hydrodynamic mode, such as large amount of calculation, slow calculation speed, and dependence on large supercomputers for computing power. It can quickly calculate the spatial distribution of pollutant concentration field and realize rapid twin mapping and visualization of geographic processes; while sensor monitoring data can only represent the "instantaneous" state of water environment elements at a certain monitoring point, and it is difficult to reflect the large-scale spatial distribution changes of geodynamic processes. The present invention can quickly calculate the spatial distribution field data of island, reef and lagoon water environment at different times in the future based on real-time monitoring data and using lightweight deep learning models.

[0044] The above description is only a specific implementation of the present invention, but the protection scope of the present invention is not limited thereto. Any person familiar with the technology can understand and think of any changes or substitutions within the technical scope disclosed by the present invention, which should be included in the scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0045] Figure 3 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention is shown in FIG. Figure 3 As shown, the device includes a processor 60, a memory 61, an input device 62 and an output device 63; the number of processors 60 in the device can be one or more. Figure 3 A processor 60 is taken as an example; the processor 60, the memory 61, the input device 62 and the output device 63 in the device can be connected by a bus or other means. Figure 3 The example of connecting through bus is taken in the following.

[0046] The memory 61, as a computer-readable storage medium, can be used to store software programs, computer executable programs and modules, such as program instructions / modules corresponding to the twin mapping method of the fusion power mode and the deep learning model in the embodiment of the present invention. The processor 60 executes various functional applications and data processing of the device by running the software programs, instructions and modules stored in the memory 61, that is, realizing the above-mentioned twin mapping method of the fusion power mode and the deep learning model.

[0047] The memory 61 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and at least one application required for a function; the data storage area may store data created according to the use of the terminal, etc. In addition, the memory 61 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 61 may further include a memory remotely arranged relative to the processor 60, and these remote memories may be connected to the device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0048] The input device 62 may be used to receive input digital or character information and generate key signal input related to user settings and function control of the device. The output device 63 may include a display device such as a display screen.

[0049] An embodiment of the present invention also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, it implements the twin mapping method of the fusion power mode and the deep learning model of any embodiment.

[0050] The computer storage medium of the embodiment of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, a device or a device or used in combination with it.

[0051] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0052] The program code embodied on the computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0053] Computer program code for performing the operations of the present invention may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as C or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.

Claims

1. A twin mapping method integrating power mode and deep learning model, characterized in that: include: According to the geodynamic process of the twin mapping scenario, the environmental factor data of the target area is collected in real time; Using the physical dynamic model, numerically simulate the twin mapping scene of the target area, and use the environmental factor data to correct the numerical simulation results to obtain the spatial distribution data of the environmental factor; The spatial distribution data of the environmental factors are used as samples to train a lightweight deep learning model, wherein the trained model uses the historical spatial distribution data of the environmental factors as input and uses the future spatial distribution data of the environmental factors as output; Input the environmental factor data collected in real time into the trained model, dynamically predict the future environmental factor data of the target area in real time and display them visually; The twin mapping scenario is the self-purification of pollutants in island reef lagoons; The real-time collection of environmental factor data of the target area according to the geodynamic process of the twin mapping scenario includes: collecting the water environmental factor data of the lagoon in real time, including nitrogen concentration, phosphorus concentration, seawater temperature, seawater salinity, and ocean current field, using sensors deployed in the sewage outlet area of ​​the island and reef lagoon according to the dynamic process of water exchange between the lagoon and the open sea during the self-purification of pollutants in the island and reef lagoon; The method uses the physical dynamic model to numerically simulate the twin mapping scene of the target area, and uses the environmental factor data to correct the numerical simulation results to obtain the spatial distribution data of the environmental factors, including: determining the model parameterization scheme according to the opening and closing of the island and reef lagoon, terrain and wind field effects, and using the physical dynamic model to numerically simulate the lagoon water exchange process according to the historical observation data and field survey data of the island and reef lagoon; using the environmental factor data collected in real time to perform parameter correction on the numerical simulation results to obtain the spatial distribution data of nitrogen concentration, phosphorus concentration, seawater temperature, seawater salinity, and ocean current field at different times in the lagoon; The parameters of the numerical simulation results are corrected by using the environmental factor data collected in real time to obtain the spatial distribution data of nitrogen concentration, phosphorus concentration, seawater temperature, seawater salinity, and ocean current field at different times in the lagoon, including: Interpolate the environmental factor data collected in real time and convert it into data in the same format as the numerical simulation results; The converted data are integrated with the numerical simulation results using a data assimilation method to obtain the spatial distribution data of nitrogen concentration, phosphorus concentration, seawater temperature, seawater salinity, and ocean current field at different times in the lagoon, wherein the data assimilation method includes at least one of a variational method, a Kalman filter, or a particle filter.

2. The method according to claim 1, characterized in that The model parameterization scheme is determined according to the opening and closing of the island and reef lagoon, terrain and wind field effects, and the physical dynamic model is used to numerically simulate the lagoon water exchange process based on the historical observation data and field investigation data of the island and reef lagoon, including: The large-scale sea area where the island reef lagoon is located is set as the initial area to determine the parameterization scheme for model operation; four external forcing conditions, wind, open boundary, river, and meteorological conditions, are used to run the large-scale coarse-resolution FVCOM hydrodynamic model to simulate the large-scale field data of water environment factors; According to the target island and reef lagoon coastline, topography and geomorphology data, the island and reef lagoon area is refined and network nested to construct the high-precision FVCOM hydrodynamic model of the island and reef lagoon area; The output results of the large-scale coarse-resolution FVCOM hydrodynamic model are used as the open boundary conditions of the water environment factor data. Combined with the observation data of water environment factors in the historical period of island and reef lagoons and the marine dynamic environment survey data, the FVCOM high-precision hydrodynamic model of the island and reef lagoon area is called to simulate and generate the water environment data of the lagoon in the historical period.

3. The method according to claim 1, characterized in that The lightweight deep learning model is a convolutional-long short-term memory neural network model.

4. The method according to claim 1, characterized in that: The step of training the lightweight deep learning model using the spatial distribution data of the environmental factors as samples includes: Extract the spatial distribution data of environmental factors at multiple consecutive moments before and after any moment as samples; The spatial distribution data of the environmental factor at multiple consecutive moments before the moment is input into a lightweight deep learning model for training, so that the model output continuously approaches the spatial distribution data of the environmental factor at multiple consecutive moments after the moment.

5. The method according to claim 4, characterized in that The input and output of the model are both four-dimensional data, and the dimensions are time, number of environmental factors, and grid size.

6. The method according to claim 1, characterized in that The real-time dynamic prediction of the future environmental factor data of the target area and the visual display thereof include: Extract environmental factor data for consecutive time periods; The grid field data in the environmental factor data are visualized in a hierarchical manner using progressive colors in chronological order, wherein the grid field data include nitrogen concentration, phosphorus concentration, seawater temperature and seawater salinity; Perform particle flow visualization on ocean current field data; The ocean current field data visualization layer is superimposed on the grid field data visualization layer to realize dynamic visualization of nitrogen concentration, phosphorus concentration, seawater temperature and seawater salinity in the lagoon as they change with the ocean current field.

7. An electronic device, characterized in that: include: one or more processors; a memory for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the twin mapping method of the fusion power mode and the deep learning model as described in any one of claims 1-6.

8. A computer-readable storage medium, characterized in that: A computer program is stored thereon, which, when executed by a processor, implements the twin mapping method of the fusion power mode and the deep learning model as described in any one of claims 1-6.

Images